Biotechnol. Prog.

2005, 21, 621626

621

A Generalized Model for Settling Velocity of Aerobic Granular Sludge

Yu Liu,* Zhi-Wu Wang, Yong-Qiang Liu, Lei Qin, and Joo-Hwa Tay
Division of Environmental and Water Resources Engineering, School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798

Aerobic granulation is a novel biotechnology recently receiving intensive research attention. Aerobic granules developed in SBR can be as big as several millimeters, thus the traditional models describing the settling velocity of activated sludge are no long valid in aerobic granules culture. In this study, a new type of model was developed for the settling velocity of aerobic granules. This model shows that the settling velocity of aerobic granules is the function of SVI, mean size of granules and biomass concentration of granules. When the size of bioparticle is small enough, the proposed model reduces to the well-known Vesilind equation. Results indicated that the proposed model could satisfactorily fit experimental results obtained in the course of aerobic granulation under different conditions, while the Vesilind equation failed to or very poorly fit the experimental data. In addition, the proposed model can also be extended to anaerobic granules. The settling velocity is one of the most important parameters in both aerobic and anaerobic granulation, and successful biogranulation is highly related to the manipulation of settling velocity. It was demonstrated that the proposed model can sever as a useful tool for design and operation engineers to properly select the settling velocity for enhanced aerobic and anaerobic granulation.

Introduction
Aerobic granulation is a novel biological process recently developed for wastewater treatment. Basically, aerobic granules are formed through cell-to-cell immobilization without supply of carriers. Aerobic granules can be differentiated from activated sludge by their spherical shape, large size in the range of 0.2-9 mm, good settleability, and high biomass retention. Thus far, aerobic granulation technology has been tested for the treatment of a very wide spectrum of wastewaters containing organics, nitrogen, and phosphorus (Morgenroth et al., 1997; Tay et al., 2001; Lin et al., 2003; Yang et al., 2003). In conventional activated sludge processes, biosolidliquid separation is one of the most critical operations and determines the quality of effluent from the secondary clarifier as well as subsequent sludge dewatering. Poor biomass separation is a very common problem encountered in the secondary settling tank (Higgins and Novak, 1997; Droste, 1997; Jin et al., 2003). Zone settling velocity is a well-known parameter that can be used to describe sludge settleability, and the most popular model for predicting the settling velocity of activated sludge is proposed by Vesilind (1968) in the form of:

equately and accurately describe the settling velocity of activated sludge with different sludge volume indices (SVI). To incorporate the effect of SVI on the measured settling velocity, many modified formula have since been proposed (Akca et al., 1993; Renko, 1998; Bye and Dold, 1999). It should be realized that in the Vesilind and those modified equations, the contribution of floc size to the settling velocity is commonly ignored. These in turn imply that the existing velocity equations developed for activated sludge could no longer be valid and would be inapplicable for aerobic granules as well as for anaerobic granules with larger sizes. Thus, this study attempted to develop a generalized equation for settling velocity of aerobic granules, which would be also applicable for anaerobic granules. This study is expected to provide useful information for the design of a granular sludge system for wastewater treatment.

Materials and Methods

Three column sequencing batch reactors (SBR) (R1 to R3), each with a working height of 1 m and an inner diameter of 5 cm, were operated at respective exchange ratios of 60, 40, and 20%. The initial chemical oxygen demand (COD) concentration was maintained at 1000 mg L-1 in R1 to R3. The operation details of the SBRs can be found elsewhere (Tay et al., 2001). Aeration was supplied from the bottom of each reactor at a rate of 3 l min-1. R1 to R3 were seeded by 500 mL of activated sludge acclimatized to sodium acetate. Detail composition of the synthetic substrate used in this study can be found elsewhere (Moy et al., 2002). Sludge volume index (SVI) and suspend solids (SS) were measured by standard methods (APHA, 1998). The zone settling velocities of aerobic granules were directly determined in R1 to R3 according to standard methods (APHA, 1998), and the

Vs ) Vo e-kX

(1)

in which Vs is the settling velocity, Vo is the initial settling velocity, and k is the empirical settling parameter. The Vesilind equation only accounts for the effect of biomass concentration on the settling velocity, and strong evidence shows that this equation cannot ad* Corresponding author. E-mail: cyliu@ntu.edu.sg.
10.1021/bp049674u CCC: $30.25

2005 American Chemical Society and American Institute of Chemical Engineers Published on Web 12/03/2004

622

Biotechnol. Prog., 2005, Vol. 21, No. 2

size of the granular sludge was determined by a laser particle size analysis system (Malvern Mastersizer Series 2600) or image analyzer (Quantimnet 500 Image Analyzer; Lecia Cambridge Instruments). Model Development. The zone settling velocity of suspended activated sludge has been often described by the empirical Vesilind equation (eq 1). In the study of the settling velocity of activated sludge, the role of particle size is ignored and is not quantitatively reflected in the Vesilnd equation and the modified Vesilind equations (Wahlberg and Keinath, 1988; Hartel and Po pel, 1992; Akca et al., 1993; Renko, 1998; Giokas et al., 2003). This is because the size of activated sludge flocs is very small, normally less than 100 m, and activated sludge is relatively homogeneous. However, in an aerobic granular sludge reactor, the size of aerobic granules may vary in a very wide range of 0.25-9 mm. In this case, the effect of the granule size on Vs must be taken into account. According to the well-known Stokes formula, the settling velocity of a particle can be estimated as follows:

in which R is a constant coefficient and equals g/18a. Equation 8 shows that the settling velocity of aerobic granular sludge is determined by the size of the granule, SVI, and biomass concentration of granules. Model Verification. To determine the constant involved, eq 8 is linearized as follows:

Y ) ln R - X
in which

(9)

Y ) ln

Vs SVI dp2

(10)

Vs )

g(Fp - F)dp2 18

(2)

in which Vs is settling velocity of a particle, dp is the diameter of a particle, Fp is the density of a particle, F is the density of a solution, and is the viscosity of the solution. This equation shows that the settling velocity of a particle is mainly determined by the density and diameter of aggregates in solution. In studies on aerobic granulation, SVI has been commonly measured to reflect the compactness of microbial association. According to the definition of SVI (APHA, 1998), SVI can be calculated in a way such that

Thus, plotting Y versus X gives a straight line with a slope of - and an intercept of ln R in the Y axis. Changes in size, biomass concentration, SVI, and corresponding settling velocity of microbial aggregates were determined in the course of aerobic granulation in R1 to R3 operated at different exchange ratios as described earlier. Figures 1a-3a show that eq 8 can provide a satisfactory fitting to the experimental data obtained under different conditions, indicated by a respective correlation coefficient of 0.94, 0.89, and 0.88. For comparison purposes, the Vesilind equation (eq 1) is also used to fit the experimental results as presented in Figures 1a-3a. To do so, eq 1 is expressed as follows:

ln Vs ) ln Vo - kX

(11)

SVI )

settled sludge volume dry weight of suspended solids

(3)

It is reasonable to consider that the settled sludge volume can be regarded approximately as the wet sludge volume, and rearrangement of eq 3 yields

1 dry weight of suspended solids ) SVI wet sludge volume

(4)

In fact, the term 1/SVI represents the dry density of suspended solids that is equal to (Fp - F), i.e.

Fp - F )

1 SVI

(5)

Equation 11 shows that ln Vs versus X gives a straight line with a slope of -k and an intercept of ln Vo in the Y axis. The fittings of the Vesilind equation to the experimental data are presented in Figures 1b-3b. It appears from Figure 1b that the Vesilind equation fails to fit those experimental data because a positive slope is obtained in this figure, while Figures 2b and 3b show that the Vesilind equation poorly predicts the experimental data, indicated by the low correlation coefficient of 0.086 and 0.62, respectively. These in turn imply that the Vesilind equation is no longer applicable for predicting the settling velocity of aerobic granular sludge with variable and large sizes. In a recent study, Moon et al. (2003) also reported that the zone settling velocity of activated sludge was related to its size. To further verify the proposed equation (eq 8), additional experiments were conducted with aerobic granules having defined mean size in the range of 0.23-2.4 mm, while the biomass concentration was kept constant at 1.2 kg m-3. In this case, eq 8 reduces to

Substitution of eq 5 into eq 2 gives

2 g dp Vs ) 18 SVI

dp2 Vs ) SVI (6)

in which

(12)

Accumulated evidence shows that the viscosity of mixed liquors in a biological reactor is an exponential function of the concentration of biomass (Manem and Sanderson, 1996; Hasar et al., 2004), i.e.

) R e-X

(13)

) a eX

(7)

in which X is the biomass concentration, and a and are constant coefficients. Combining eqs 6 and 7 yields

Figure 4 shows a plot of dp2/SVI against Vs. A good agreement between the eq 12 prediction and the experimental data is obtained, indicated by a correlation coefficient of 0.94. Figure 4 further confirms eq 8, and also implies that the Vesilind equation cannot be applied to microbial granules culture.

dp2 -X e Vs ) R SVI

Discussion

(8)

It appears from Figures 1-4 that the proposed model can offer a satisfactory prediction to the settling velocities

Biotechnol. Prog., 2005, Vol. 21, No. 2

623

Figure 1. Settling velocity of microbial aggregates as the function of SVI, mean size, and biomass concentration observed in the course of aerobic granulation in SBR operated at an exchange ratio of 60%. (a) Equation 8 prediction with respective R and values of 5.94 107 m2 kg-1 h-1 and 0.98 m3 kg-1, and correlation coefficient ) 0.94. (b) Failure in prediction by the Vesilind equation with a positive slope.

Figure 3. Settling velocity of microbial aggregates as the function of SVI, mean size, and biomass concentration observed in the course of aerobic granulation in SBR operated at an exchange ratio of 40%. (a) Equation 8 prediction is shown by solid curve with respective R and values of 1.95 107 m2 kg-1 h-1 and 0.77 m3 kg-1, and correlation coefficient ) 0.88. (b) Prediction by the Vesilind equation is shown by dotted curve with respective Vo and k values of 4.43 m h-1 and 0.142 m3 kg-1, and correlation coefficient ) 0.62.

Figure 4. Settling velocity of aerobic granules as the function of SVI, mean size of aerobic granules at a constant biomass concentration observed in the course of aerobic granulation in SBR operated at an exchange ratio of 40%. Equation 8 prediction is shown by solid curve with value of 3.4 105 m2 kg-1 h-1.

Figure 2. Settling velocity of microbial aggregates as the function of SVI, mean size, and biomass concentration observed in the course of aerobic granulation in SBR operated at an exchange ratio of 40%. (a) Equation 8 prediction is shown by solid curve with respective R and values of 2.19 107 m2 kg-1 h-1 and 0.81 m3 kg-1, and correlation coefficient ) 0.89. (b) Prediction by the Vesilind equation is shown by dotted curve with respective Vo and k values of 7.87 m h-1 and 0.0186 m3 kg-1, and correlation coefficient ) 0.086.

obtained. These may indicate that for aerobic granules with large sizes, the Vesilind equation (originally developed for activated sludge) is no longer valid. In the activated sludge process, the size of activated sludge flocs is usually thought to be uniform without significant changes, and if experiments are conducted with the same activated sludge, for example, the SVI can also be regarded as constant. In this case, eq 8 reduces to
-X V s ) V o e

(14)

in which V o is equivalent to Vo in the Vesilind equation and is defined as follows:

determined in the course of aerobic granulation, while the Vesilind equation as commonly used in the activated sludge process is very poor in describing the data

dp2 V o ) R SVI

(15)

624

Biotechnol. Prog., 2005, Vol. 21, No. 2

Obviously, eq 14 can be regarded as the Vesilind equation (eq 1), while eq 15 probably for the first time reveals that V o in eq 14 or Vo in the Vesilind equation may not be a true constant, but instead is a function of SVI and the mean size of bioflocs. Previous studies on activated sludge compared and evaluated the settling velocity parameters based on SVI, and a number of the empirical formula correlating V o and SVI had been proposed, such as eq 16 by Wahlberg and Keinath (1988), Hartel and Po pel (1992), and Ozinsky and Ekamma (1995), eq 17 by Akca et al. (1993):

Vo ) m1 e-m2SVI
and

(16)

Vo ) n1SVI-n2

(17)
Figure 5. Three-dimensional visualization of eq 15 with R ) 5.94 107 m2 kg-1 h-1.

in which m1, m2, n1, and n2 are positive empirical constants. It should be realized that the effect of sludge size on Vo is not accounted for in eqs 16 and 17, which are completely empirical and are generated from curve fitting. Using Matlab6.5, Figure 5 further visualizes the effects of SVI in a range of 30-100 mL g-1 and dp varying from 0 to 5 mm on V o. Two distinct zones can be seen in Figure 5: (i) for bioparticles with a mean size less than 0.3 mm (which indeed is in the size range of activated sludge flocs), it is found that V o is not sensitive to changes in both SVI and dp, and could be regarded as constant. These may imply that the Vesilind equation is validated for activated sludge with a typical mean size less than 0.3 mm, and are in good agreement with Bye and Dold (1998) noting that it was questionable to correlate of Vo or V o to SVI for activated sludge; (ii) for bioparticles with a mean size much larger than 0.3 mm, both SVI and dp have significant effect on V o. Consequently, eq 8 is superior to the existing formula for setting velocity. The contribution of particle size to the settling velocity is taken into account in eq 8, which is extremely important for aerobic granular sludge with a variable and larger size than conventional activated sludge bioflocs. Recently, more and more evidence shows that aerobic granulation is a microbial process driven by selection pressures, and among all selection pressures known thus far, the settling time of sludge in SBR is the strongest one (Qin et al., 2004). No matter how other operating parameters are manipulated, aerobic granulation would fail in SBR if the settling time is not properly controlled, i.e., settling time must be set as short as possible to enhance and accelerate the aerobic granulation process (Qin et al., 2004). In fact, a short settling time was preferentially selected for the growth of good settling bacteria, and the sludge with a poor settleability would be washed out. The question remaining unanswered thus far is how to determine the critical settling time for a rapid aerobic granulation. In a column SBR for aerobic granulation, the effluent is discharged from the discharging port (Figure 6). If the distance for water to travel to the discharge port is L, the corresponding travelling time can be calculated as

Figure 6. Geometry of a column SBR.

washed out of the reactor. Thus, the critical setting time for aerobic granulation in SBR would be

critical settling time )

Substitution of eq 8 into eq 19 gives

L Vs

(19)

critical settling time )

L SVI eX R d2
p

(20)

travelling time to the discharge port )

L (18) Vs

This equation shows that the higher Vs results in a shorter travelling time of particles to the discharge port, i.e., the bioparticles with a travelling time longer than the settling time chosen for the SBR system will be

It can be seen that the critical settling time for the enhanced aerobic granulation is a function of SVI, mean size of bioparticles, and biomass concentration. Equation 20 for the first time provides theoretical guidelines for selecting and adjusting the settling time in a timely manner according to the parameters describing settleability in terms of particle size and SVI. In fact, this strategy has been successfully applied to accelerate aerobic granulation in SBR by adjusting the settling time according to changes in the settleability of bioparticles (Qin et al., 2004). In the family of biogranulation, another key member is so-called anaerobic granulation which has been commonly reported in upflow anaerobic sludge blanket (UASB) reactor. Physically, anaerobic granules are similar to aerobic granules in terms of size, SVI, and so on. Show et al. (2004) studied anaerobic granulation in a labscale UASB reactor, and measured SVI, size, biomass

Biotechnol. Prog., 2005, Vol. 21, No. 2

625
Alphenaar, P. A.; Visser, A.; Lettinga, G. The effect of liquid upflow velocity and hydraulic retention time on granulation in UASB reactors treating wastewater with a high-sulphate content. Bioresour. Technol. 1993, 43, 249-258. APHA. Standard methods for the examination of water and wastewater, 20th ed.; American Public Health Association: Washington, DC, 1998. Bye, C. M.; Dold, P. L. Evaluation of correlations for zone settling velocity parameters based on sludge volume indextype measures and consequences in settling tank design. Water Environ. Res. 1999, 71, 1333-1344. Droste, R. L. Theory and practice of water and wastewater treatment; John Wiley: New York, 1997. Giokas, D. L.; Daigger, G. T.; Sperling, M.; Kim, Y.; Paraskevas, P. A. Comparison and evaluation of empirical zone settling velocity parameters based on sludge volume index using a unified settling characteristics database. Water Res. 2003, 37, 3821-3836. Hartel, L.; Po pel, J. H. A dynamic secondary clarifier model including processes of sludge thickening. Water Sci. Technol. 1992, 25, 267-284. Hasar, H.; Kinaci, C.; U nlu , A.; Tog rul, H.; Ipek, U. Rheological properties of activated sludge in a sMBR. Biochem. Eng. J. 2004, 20, 1-6. Higgins, M.; Novak, J. Dewatering and settling of activated sludges: the case for using cation analysis. Water Environ. Res. 1997, 69, 225-232. Hulshoff Pol, L. W.; Heijnekamp, K.; Lettinga, G. The selection pressure as a driving force behind the granulation of anaerobic sludge. In Granular Anaerobic Sludge: Microbiology and Technology; Lettinga, G., Zehnder, A. J. B., Eds.; Grotenhuis, J. T. C., and Hulshoff Pol, L. W.: Wageningen, The Netherlands, 1988; pp 153-161. Jin, B.; Wilen, B. M.; Lant, P. A comprehensive insight into floc characteristics and their impact on compressibility and settleability of activated sludge. Chem. Eng. J. 2003, 95, 221234. Lin, Y. M.; Liu, Y.; Tay, J. H. Development and characteristics of phosphorous-accumulating microbial granules in sequencing batch reactors. Appl. Microbiol. Biotechnol. 2003, 62, 430-435. Liu, Y.; Xu, H. L.; Yang, S. F.; Tay, J. H. The mechanisms and models for anaerobic granulation in upflow anaerobic sludge blanket reactor. Water Res. 2003, 37, 661-673. Manem, J.; Sanderson, R. Membrane bioreactor. In Water treatment membrane processed; Mallevialle, J., Odendaal, P. E., Wiesner, M. R., Eds.; McGraw-Hill: New York, 1996; pp 17.1-17.31. Moon, B. H.; Seo, G. T.; Lee, T. S.; Kim, S. S.; Yoon, C. H. Effects of salt concentration on floc characteristics and pollutants removal efficiencies in treatment of seafood wastewater by SBR. Water Sci. Technol. 2003, 47, 65-70. Morgenroth, E.; Sherden, T.; van Loosdrecht, M. C. M.; Heijnen, J. J.; Wilderer, P. A. Aerobic granular sludge in a sequencing batch reactor. Water Res. 1997, 31, 3191-3194. Moy, B. Y. P.; Tay, J. H.; Toh, S. K.; Liu, Y.; Tay, S. T. L. High organic loading influences the physical characteristics of aerobic granules. Lett. Appl. Microbiol. 2002, 34, 407412. Ozinsky, A. E.; Ekama, G. A. Secondary settling tank modelling and design Part 1: review of theoretical and practical developments. Water SA 1995, 21, 325-332. Qin, L.; Liu, Y.; Tay, J. H. Effect of settling time on aerobic granulation in sequencing batch reactor. Biochem. Eng. J. 2004, 21, 47-52. Renko, E. K. Modelling hindered batch settling Part I: A model for linking, zone settling velocity and stirred sludge volume index. Water SA 1998, 24, 325-330. Show, K. Y.; Wang, Y.; Foong, S. F.; Tay, J. H. Accelerated startup and enhanced granulation in upflow anaerobic sludge blanket reactors. Water Res. 2004, 38, 2293-2304.

Figure 7. Settling velocity of anaerobic granules as the functions of SVI, mean size, and biomass concentration observed in a lab scale UASB reactor. Data from Show et al. (2004). (a) Equation 6 prediction is shown by solid curve with respective R and values of 1.55 107 m2 kg-1 h-1 and 0.069 m3 kg-1, and correlation coefficient ) 0.94. (b) Failure in prediction by the Vesilind equation with a positive slope.

concentration, and corresponding settling velocity. Figure 7a shows that prediction by eq 8 is in good agreement with the experimental data, while the Vesilind equation once again fails to do so, indicated by a positive slope (Figure 7b). Thus, it seems reasonable to consider that eq 8 can be extended to anaerobic granulation. The UASB reactor is operated in a continuous mode, and it has been believed that anaerobic granulation in the UASB reactor is driven by hydraulic selection pressure in terms of the liquid upflow velocity, Vup (Hulshoff Pol et al., 1988; Alphenaar et al., 1993; Liu et al., 2003). The liquid upflow velocity has double-side effect on anaerobic granulation, i.e., low Vup does not encourage or enhance anaerobic granulation, while if Vup is too high, sludge washout from the UASB reactor could occur and eventually would lead to failure of anaerobic granulation. Theoretically, the critical liquid upflow velocity that can be applied to UASB reactor should be equal to Vs as predicted by eq 8. In practice, to maintain the long-term stability of UASB reactor and to avoid eventual washout of anaerobic granular sludge blanket, the liquid upflow velocity applied to the UASB reactor must be much lower than Vs. Thus, information on the settling velocity is critical in the determination of the allowable upflow velocity of the anaerobic granular sludge in UASB reactors. In this aspect, eq 8 is useful tool that may offer information related to the control of the liquid upflow velocity in UASB reactor.

References and Notes

Akca, L.; Kinaci, C.; Karpuzcu, M. A model for optimum design of activated sludge plants. Water Res. 1993, 27, 1461-1468.

626
Tay, J. H.; Liu, Q. S.; Liu, Y. Microscopic observation of aerobic granulation in sequential aerobic sludge reactor. J. Appl. Microbiol. 2001, 91, 168-175. Vesilind, P. A. Design of prototype thickeners from batch settling tests. Water Sewage Works 1968, 115, 302-307. Wahlberg, E. J.; Keinath, T. M. Development of settling flux curves using SVI. J. Water Pollut. Control Fed. 1988, 60, 2095-2100.

Biotechnol. Prog., 2005, Vol. 21, No. 2

Yang, S. F.; Tay, J. H.; Liu, Y.A novel granular sludge sequencing batch reactor for removal of organic and nitrogen from wastewater. J. Biotechnol. 2003, 106, 77-86.

Accepted for publication October 14, 2004. BP049674U